Accelerating scientific discovery with Co-Scientist

2026-05-19 · Source: Machine learning : nature.com subject feeds · Field: Science & Research — Artificial Intelligence & Machine Learning, Health & Medical Research, Research Methodology & Innovation · Depth: Expert, quick

Summary

Co-Scientist is a multi-agent AI system, built on Gemini, designed to accelerate scientific discovery by augmenting hypothesis generation. This system helps scientists formulate demonstrably novel research hypotheses for experimental verification, conditioned on research objectives and prior evidence. Its architecture features a multi-agent design with an asynchronous task execution framework, enabling flexible compute scaling, and incorporates a tournament evolution process for self-improving hypothesis generation. Automated evaluations confirm that scaling test-time compute continuously improves hypothesis quality. While general-purpose, Co-Scientist was validated in three biomedical applications: drug repurposing, novel target discovery, and explaining anti-microbial resistance. Notably, it identified new drug repurposing candidates and synergistic combination therapies for acute myeloid leukemia, which were subsequently validated through in vitro experiments. This work was published on May 19, 2026.

Key takeaway

For research scientists exploring complex problems, Co-Scientist offers a powerful AI-driven approach to accelerate discovery. You should consider integrating multi-agent AI systems like Co-Scientist into your workflow to generate and validate novel hypotheses more efficiently. This system's ability to identify new drug candidates and synergistic therapies, validated in vitro, suggests a significant shift in how you might approach early-stage research and experimental design.

Key insights

Co-Scientist, a Gemini-based multi-agent AI, accelerates scientific discovery by generating and refining novel, verifiable hypotheses.

Principles

• Multi-agent AI enhances hypothesis generation.
• Asynchronous task execution scales compute.
• Tournament evolution improves hypothesis quality.

Method

Co-Scientist agents continuously generate, critique, and refine hypotheses, leveraging asynchronous task execution and a tournament evolution process, accelerated by scaling test-time compute.

In practice

• Identify drug repurposing candidates.
• Discover novel therapeutic targets.
• Explain anti-microbial resistance mechanisms.

Topics

• Multi-agent AI
• Gemini
• Hypothesis Generation
• Drug Repurposing
• Biomedical Research
• Scientific Discovery

Best for: AI Scientist, Research Scientist

Related on AIssential

• Co-Scientist: A multi-agent AI partner to accelerate research — Co-Scientist is a multi-agent AI system that iteratively generates, debates, and refines scientific hypotheses to accelerate discovery.
• Generating novel scientific hypotheses with Co-Scientist — Co-Scientist is a multi-agent AI system designed to accelerate scientific discovery by automating literature review and hypothesis generation.
• 12 AI Co-Scientists of 2026 — AI co-scientists are rapidly accelerating scientific discovery across diverse fields through autonomous hypothesis generation, experimentation, and analysis.
• 4 ways researchers are collaborating with Co-Scientist to solve big problems — Co-Scientist is a multi-agent AI system accelerating scientific hypothesis generation and refinement across life sciences.
• Advancing Scientific Discovery in the Agentic AI Era — Accelerated computing and AI agents are dramatically speeding up scientific discovery beyond human processing limits.
• Gemini for Science: AI experiments and tools for a new era of discovery — Gemini for Science leverages AI agents to automate complex research tasks, significantly accelerating scientific discovery and exploration.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine learning : nature.com subject feeds.